International Mathematical Forum, Vol. 9, 2014, no. 15, 725 - 731 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/imf.2014.4344

Convergence Theorems for Pettis of Functions Taking Values in Locally Convex Spaces

Zamir Selko

Faculty of Natural Sciences Department of Mathematics University of Elbasan, Albania

Copyright c 2014 Zamir Selko. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

In this paper we present convergence theorems for Pettis integral of functions defined on a complete probability space and taking values in a complete locally convex .

Mathematics Subject Classification: 28B05, 46G10

Keywords: Pettis integral, convergence theorems, complete locally convex topological vector space

1 Introduction

We present convergence theorems for Pettis integral of functions defined on a complete probability space and taking values in a complete locally convex topological vector space, Theorems 2.2 and 2.3, which are the analogues of the known convergence theorems for Pettis integral of functions taking values in a . For that purpose, we apply a technique which is based on the limit projective of Banach spaces, Lemma 2.1. Theorem 2.2 is the analogue of Vitali Theorem for Pettis integral of func- tions taking values in a Banach space, Theorem 8.1 in [6]. Theorem 2.2 has been proved by another approach in [7]. Theorem 2.3 is the analogue of The- orem 2.8 in [9]. For more information on convergence theorems for Pettis integral, see [6], [7], [8], [4] and [2]. 726 Zamir Selko

Throughout this paper (Ω, Σ,μ) is a complete probability space and V is a complete locally convex space with its topology τ and topological dual V .By P we denote the family of all continuous semi-norms in this space; for every p ∈ P, V p denotes the quotient vector space of the vector space V with respect  p to the equivalence relation x ∼p y ⇔ p(x − y) = 0; the map φp : V → V is the canonical quotient map, thus is the equivalence class of an element with  p respect to the relation ”∼p”; the quotient normed space (V ,p) is called the normed component of the space V , where p(φp(x)) = p(x), for each x ∈ V ; p the Banach space (V , p) which is the completion of the space (V p,p) is called  the Banach component of the space V ; Vp and V p are the topological duals of p (V p,p) and (V , p) respectively. It is easy to see that

 V = {vp ◦ φp/vp ∈ Vp ,p∈ P}, (1) because for every v ∈ V , we have that |v(.)|∈P . For every p, q ∈ P  q  p such that p ≤ q, we define the map gpq : V → V by gpq(wq)=wp, for all  q wq ∈ V , where wp = φp(x), for some vector x ∈ wq. We define also the map q p gpq : V → V as the continuous linear extension of gpq, for every p, q ∈ P such that p ≤ q. The following definition is given by [1], Definition 1.

Definition 1.1 A function f : S → V is called Pettis integrable if the function υ ◦f is Lebesgue integrable for each υ ∈ V and if for every measurable set E of Ω, there is a vector xE ∈ V such that υ (xE)= υ (f(s)) for every υ ∈ V . The vector xE is said to be Pettis integral of function f on E and we denote: 

xE =(P ) f. E

If V is a Banach space, this definition is the same with Definition 2, [3], p.52.

Definition 1.2 A family H of real-valued integrable functions defined on Ω is said to be uniformly integrable if it satisfies the conditions  | | ∞ 1. suph∈H Ω h < ,

2. For each >0 there is δ() > 0 such that the inequality  sup |h| < h∈H E

holds for every E ∈ Σ satisfying μ(E) ≤ δ(). Convergence theorems 727

Let f :Ω→ V be a function at let p ∈ P. We put

p  Zf = {vp ◦ (φp ◦ f)/vp ∈ B(Vp )},

  where B(Vp ) is the closed unit ball in Vp . It is clear to see that

p Zf = {vp ◦ (φp ◦ f)/vp ∈ B(V p)}

where B(V p) is the closed unit ball in V p.

2 Convergence theorems for Pettis integral

The proof of the following auxiliary lemma is similar in spirit to the proof of Lemma 4.1 in [5].

Lemma 2.1 Let f :Ω→ V be a function. The function f is Pettis inte- grable if and only if for every p ∈ P the function φp ◦ f is Pettis integrable in p the Banach component (V , p). In this case, we have  

φp((P ) f)=(P ) φp ◦ f for all p ∈ P,E∈ Σ. (2) E E

Proof. Suppose that the function f is Pettis integrable and let p ∈ P. Then, by Definition 1.1 and (1), the function vp ◦ (φp ◦ f)=v ◦ f is the Lebesgue  integrable, for each vp ∈ B(Vp ). We have also that for each E ∈ Σ, there is a ∈ ◦ ∈ vector xE V such that v (xE)= E v f, for all v V . Hence, we obtain by (1) that   (vp ◦ φp)(xE)= (vp ◦ φp) ◦ f for each vp ∈ B(Vp ). E

Therefore, the function φp ◦ f is Pettis integrable in the Banach component p ◦ ∈ (V , p) and φp(xE)=(P ) E φp f, for all E Σ. Conversely, assume that for every p ∈ P the function φp ◦ f is Pettis p integrable in the Banach component (V , p), and let v ∈ V and E ∈ Σ  are given. Then, by the equality (1), there exist p ∈ P and vp ∈ B(Vp ) such that v = vp ◦ φp and the function v ◦ f = vp ◦ (φp ◦ f) is Lebesgue integrable. p We have also that for each p ∈ P there exists Ip(E) ∈ V such that   vp(Ip(E)) = vp ◦ (φp ◦ f)= vp ◦ (φp ◦ f) for all vp ∈ V p (3) E E

 | p where vp = vp V . 728 Zamir Selko

Assume that two arbitrary continuous seminorms p and q such that p ≤ q n  q q are given. Since Iq(E) ∈ V , there is a sequence (wq ) ⊂ V such that n n n n limn→∞ wq = Eq(E). Then, limn→∞ wp = gpq(Iq(E)), where wp = gpq(wq ), for all n ∈ N. Hence

n lim vp(wp )=vp(gpq(Iq(E))) for all vp ∈ V p, (4) n→∞  | p where vp = vp V . By virtue of (3), we have  n n lim vp(wp ) = lim (vp ◦ gpq)(wq )= (υp ◦ gpq) ◦ (φq ◦ f) n→∞ n→∞ E = υp ◦ (gpq ◦ φq) ◦ f = υp ◦ (φp ◦ f)=υp(Ip(E)) E E

∈  | p for all vp V p, where vp = vp V . The last result together with (4) yields

υp(gpq(Iq(E))) = υp(Ip) for all vp ∈ V p.

Therefore by Corollary IV.6.2 in [11], we obtain gpq(Iq(E)) = Ip(E). Hence, by Theorem II.5.4 in [10], there is a vector If (E) ∈ V such that

φp(If (E)) = Ip(E) for each p ∈ P.

Thus, since v = vp ◦ φp, we have   v (If (E)) = vp(φp(If (E))) = υp(Ip(E)) = υp ◦ (φp ◦ f)= (υp ◦ φp) ◦ f. E E Hence  v (If (E)) = v ◦ f, E and since v and E are arbitrary, the last equality holds for each v ∈ V and ∈ E Σ. This means that f is Pettis integrable and If (E)=(P ) E f, for all E ∈ Σ. Hence, the equality (2) holds and the proof is finished.

Theorem 2.2 Let (fn) be a sequence of Pettis integrable functions fn :Ω→ V and let f :Ω→ V be a function such that (i) for every v ∈ V , we have v ◦ fn → v ◦ fμ- a.e.;  p (ii) for every p ∈ P, we have Zf is uniformly integrable. n∈Æ n Then, f is Pettis integrable and, for each E ∈ Ω, we have  

lim (P ) fn =(P ) f n→∞ E E in the σ(V,V ). Convergence theorems 729

Proof. By Lemma 2.1, each function φp ◦ fn is Pettis integrable in the Banach p component (V , p) and  

(P ) φp ◦ fn = φp((P ) fn) for each E ∈ Ω. E E

 By hypothesis and (1), for each vp ∈ Vp , we have υp ◦ (φp ◦ fn) → υp ◦ (φp ◦ f), μ- a.e. Thus, the conditions of Theorem 8.1 in [6] are satisfied. Therefore, the p function φp ◦ f is Pettis integrable in the Banach component (V , p) and   lim υp((P ) φp ◦ fn)=υp((P ) φp ◦ f) for each E ∈ Ω, vp ∈ V p. n→∞ E E Hence, by Lemma 2.1, we obtain that f is Pettis integrable and    lim (vp ◦ φp)((P ) fn)=(vp ◦ φp)((P ) f) for all E ∈ Ω, vp ∈ Vp , n→∞ E E

 | p where vp = vp V . Further, we obtain by (1) that   lim v ((P ) fn)=v ((P ) f) for all E ∈ Ω,v∈ V n→∞ E E and the proof is finished.

Theorem 2.3 Let (fn) be a sequence of Pettis integrable functions fn :Ω→ V converging point-wise to a function f :Ω→ V in (V,τ). Then the following are equivalent.  ∈ p (i) for every p P, the family n∈N Zfn is uniformly integrable; (ii) the function f is Pettis integrable and, for each E ∈ Ω, we have  

lim (P ) (fn)=(P ) f in (V,τ). n→∞ E E

Proof. Note that the sequence (fn) converges point-wise to f in (V,τ), if and only if for every p ∈ P the sequence φp ◦ fn converges point-wise to φp ◦ f in the normed component (V p, p). (i) ⇒ (ii) Let p be an element of P. We have that the sequence φp ◦ fn p converges point-wise to φp ◦ f in the Banach component (V , p). we have also by Lemma 2.1 that each function φp ◦ fn is Pettis integrable in the Banach p component (V , p) and    p (P ) (φp ◦ fn)=φp((P ) fn) ∈ V for all E ∈ Ω. (5) E E 730 Zamir Selko

So, the conditions of Theorem 2.8 in [9] are satisfied. Therefore, the function p φp ◦ f is Pettis integrable in the Banach component (V , p) and the equality  

lim (P ) φp ◦ fn =(P ) φp ◦ f, (6) n→∞ E E p holds in (V , p), for each E ∈ Ω. Let E be an element of Σ. We have that for every p ∈ P, the function φp ◦f p is Pettis integrable in the Banach component (V , p). Therefore, by Lemma 2.1, the function f is Pettis integrable and    p (P ) φp ◦ f = φp((P ) f) ∈ V . (7) E E Hence, inserting the right-hand-sides of (5) and (7) to (6), we obtain  

lim φp((P ) fn)=φp((P ) f) for each p ∈ P. n→∞ E E   Therefore, limn→∞(P ) E fn =(P ) E f in (V,τ). (ii) ⇒ (i) Assume that (ii) holds and let E ∈ Σ and p ∈ P. We have that the equality  

lim φp((P ) fn)=φp((P ) f), (8) n→∞ E E p holds in the Banach space (V , p). According to Lemma 2.1, the functions p φp ◦ f and φp ◦ fn are Pettis integrable in (V , p) and    

φp((P ) f)=(P ) φp ◦ fφp((P ) fn)=(P ) φp ◦ fn. E E E E Hence, inserting the right-hand-sides of these equalities to (8), we obtain that the equality  

lim (P ) φp ◦ fn =(P ) φp ◦ f n→∞ E E p  p holds in (V , p). Therefore, by Theorem 2.8 in [9], the family n∈N Zfn is uniformly integrable and the proof is finished. Acknowledgments. Author thanks Professor Sokol Kaliaj for the proof of Lemma 2.1.

References

[1] N. D. Chakraborty and Sk . Jaker Ali, On strongly Pettis integrable func- tions in locally convex spaces, Revista Mathematica Univ.Comp. Madrid Vol.6 , no.2, (1993). Convergence theorems 731

[2] N. D. Chakraborty and T. Choudhury, Convergence theorems for Pettis integrable functions and regular methods of summability, Math.Anal.Appl. 359 (2009), 95–105.

[3] J. Distel and J. J. Uhl, Vector Measures, American Mathematical Society, Providence, (1977).

[4] R. F. Geitz, Pettis integration, Proc.Amer.Math.Soc. no.1, 82(1981), 81– 87.

[5] S. B. Kaliaj, A. D. Tato, F. D. Gumeni, Controlled convergence theorems for Henstock-Kurzweil-Pettis integral on m-dimensional compact inter- vals, Czechoslovak Math. J. 62 (137) (2012), 243-255.

[6] K. Musial, Topics in the theory of Pettis integration, Rend. Ist. Mat. Univ. Trieste 23,(1993),177–262.

[7] K. Musial, Vitali and Lebesgue convergence theorems for Pettis integral in locally convex spaces, Atti Sem. Mat. Fis. Univ. Modena, 35(1987),159– 166).

[8] K. Musial, Pettis Integral, Handbook of theory, Vol.I,II, North- Holland, Amsterdam,(2002), 531–586.

[9] J. Rodriguez, Pointwise limits of Birkhoff integrable functions, Proc. Amer. Math. Soc. no.1,137 (2009), 235–245.

[10] H. H. Schaefer, Topological Vector Spaces, Springer-Verlag New York (1971).

[11] K. Yosida, , Springer-Verlag Berlin Heidelberg (1980).

Received: January 12, 2014